[SOLVED] strange behavior with kernel-4.14.x

nobodino · 01-03-2018, 03:02 PM

Thanks for correcting the x86_64 huge kernel.

A question: has anyone remarked that the kernel-4.4.11 is really less responsive than the 4.9.x series?

Darth Vader · 01-03-2018, 03:16 PM

Quote:

Originally Posted by nobodino

A question: has anyone remarked that the kernel-4.4.11 is really less responsive than the 4.9.x series?

That problem is discussed at large in another kernel thread.

Long story short: welcome in to new age of computing, now even slower, thanks to Intel which try to protect its IMEs to be hijacked by the bad guys even via javascript.

IF you run with an pre-2009 Intel CPU or an AMD one, there is your savior: nopti added to kernel command line and a mighty reboot.

IF you run a "modern" Intel CPU, I strongly suggest you to leave as it is, or nasty things could happen. At least according with Intel.

nobodino · 01-03-2018, 03:24 PM

So the 'Kaiser' finally hit the kernel?
From 5% to 30% loss of performance, that's a 'great performance' for Intel!
Congratulations to them!

Darth Vader · 01-03-2018, 03:32 PM

Lovely, right?

Darth Vader · 01-03-2018, 03:33 PM

Quote:

Originally Posted by volkerdi

Obviously not, or the huge kernel package would have been missing. Instead, it was wrong.

And you think I would do this on purpose? Wow.

Anyway, fixed now in -current.

Thank you for the fast update!

Regnad Kcin · 01-03-2018, 03:42 PM

4.14.11 works good here in Beijing.

no slowdowns noted and some minor issues fixed by update

using 64 bit current. no 32 bit multilib these days

Petri Kaukasoina · 01-03-2018, 04:24 PM

Quote:

Originally Posted by Darth Vader

IF you run with an pre-2009 Intel CPU or an AMD one, there is your savior: nopti added to kernel command line and a mighty reboot.

IF you run a "modern" Intel CPU, I strongly suggest you to leave as it is, or nasty things could happen. At least according with Intel.

They say the original Pentium should be ok, but Pentium Pro from 1995 is vulnerable. (Though the pti fix is only in the 64 bit kernel).

Darth Vader · 01-03-2018, 04:30 PM

I noticed that in the Slackware 32-bit are no issues with KPTI, and my unique Intel CPU (from my mini-PC) runs this one.

Still, I get:

Code:

processor       : 0
vendor_id       : GenuineIntel
cpu family      : 6
model           : 23
model name      : Intel(R) Core(TM)2 Duo CPU     P8400  @ 2.26GHz
stepping        : 6
microcode       : 0x60f
cpu MHz         : 2261.424
cache size      : 3072 KB
physical id     : 0
siblings        : 2
core id         : 0
cpu cores       : 2
apicid          : 0
initial apicid  : 0
fdiv_bug        : no
f00f_bug        : no
coma_bug        : no
fpu             : yes
fpu_exception   : yes
cpuid level     : 10
wp              : yes
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe nx lm constant_tsc arch_perfmon pebs bts cpuid aperfmperf pni
dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm sse4_1 lahf_lm tpr_shadow vnmi flexpriority dtherm
bugs            : cpu_insecure
bogomips        : 4522.84
clflush size    : 64
cache_alignment : 64
address sizes   : 36 bits physical, 48 bits virtual
power management:

tazza · 01-05-2018, 11:12 PM

Quote:

Originally Posted by Regnad Kcin

4.14.11 works good here in Beijing.

no slowdowns noted and some minor issues fixed by update

using 64 bit current. no 32 bit multilib these days

64 bit current with multilib - but can confirm no slowdown for me either on the whole (7600K) according to geekbench it's marginally faster https://browser.geekbench.com/user/132977

The main difference I see is that before going to 4.14.11 (I stayed on 4.9.x) was that the CPU at idle would sit at 800Mhz with bounces up to 2Ghz.. on 4.14.11 it bounces mainly between 2Ghz up to 4.5Gz but the temps are still sitting around 36C so I don't know if the readings are getting stuffed up somehow.

khronosschoty · 01-05-2018, 11:19 PM

Can't say I notice any slowdowns with 4.14.11.

Ilgar · 01-06-2018, 05:50 AM

Quote:

Originally Posted by nobodino

A question: has anyone remarked that the kernel-4.4.11 is really less responsive than the 4.9.x series?

For normal day-to-day use of a typical desktop user, there should be no visible slowdown. Even on my humble Celeron 2-in-1 netbook I see no difference in responsiveness. PTI hits applications making many syscalls. These are typically server software, making huge amounts of file or network I/O and virtualization stuff.

BratPit · 01-06-2018, 06:54 AM

On my Slackware Desktop /Pentium Hasswell/ there is about 10% slowdown
measured by execute du command :

Quote:

time for i in {1..20}; do du -s -x /; done

Quote:

Without PTI patch

real 0m11.479s
user 0m1.284s
sys 0m10.124s

=======
with PTI patch

real 0m12.682s
user 0m1.866s
sys 0m10.754s

And this not causing any visable slowdown or responsiveness in typical Desktop usage.

So this is like difference between 3 and 3.3 GHz processor.
There is difference but with most tasks is neglible.

Ilgar · 01-06-2018, 07:40 AM

The same "du" test gives 30% slowdown on my Celeron but it is not representative for typical desktop use. Funnily enough, some Octave scripts which I need to work with seem to be even a bit faster on 4.14.11. Of course, PTI is not supposed to make any difference there anyway. I wonder if the positive impact from the recent introduction of the ORC unwinder into the kernel may even outweigh the hit from PTI on desktop workloads.